diff mbox

cgroup: WARNING in cgroup_kill_sb

Message ID 20170306215515.GH26127@htj.duckdns.org (mailing list archive)
State New, archived
Headers show

Commit Message

Tejun Heo March 6, 2017, 9:55 p.m. UTC 
Hello, Dmitry.

Can you please see whether the following patch resolves the issue?
I'm a bit nervous about it ending up in circular dependency, but I
*think* it should be okay.

Thanks.

Comments

Dmitry Vyukov March 7, 2017, 9:11 a.m. UTC | #1 
On Mon, Mar 6, 2017 at 10:55 PM, Tejun Heo <tj@kernel.org> wrote:
> Hello, Dmitry.
>
> Can you please see whether the following patch resolves the issue?
> I'm a bit nervous about it ending up in circular dependency, but I
> *think* it should be okay.
>
> Thanks.
>
> diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
> index 0125589..9c40421 100644
> --- a/kernel/cgroup/cgroup.c
> +++ b/kernel/cgroup/cgroup.c
> @@ -1820,6 +1820,8 @@ static void cgroup_kill_sb(struct super_block *sb)
>         struct kernfs_root *kf_root = kernfs_root_from_sb(sb);
>         struct cgroup_root *root = cgroup_root_from_kf(kf_root);
>
> +       mutex_lock(&cgroup_mutex);
> +
>         /*
>          * If @root doesn't have any mounts or children, start killing it.
>          * This prevents new mounts by disabling percpu_ref_tryget_live().
> @@ -1834,6 +1836,8 @@ static void cgroup_kill_sb(struct super_block *sb)
>                 percpu_ref_kill(&root->cgrp.self.refcnt);
>
>         kernfs_kill_sb(sb);
> +
> +       mutex_unlock(&cgroup_mutex);
>  }
>
>  struct file_system_type cgroup_fs_type = {



No, still happens. Please run the repro.


[  367.607496] ------------[ cut here ]------------
[  367.608012] WARNING: CPU: 1 PID: 16161 at lib/percpu-refcount.c:317
percpu_ref_kill_and_confirm+0x3ff/0x500
[  367.608019] percpu_ref_kill_and_confirm called more than once on css_release!
[  367.608019] Kernel panic - not syncing: panic_on_warn set ...
[  367.608019]
[  367.608019] CPU: 1 PID: 16161 Comm: a.out Not tainted 4.11.0-rc1+ #311
[  367.608019] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS Bochs 01/01/2011
[  367.608019] Call Trace:
[  367.608019]  dump_stack+0x2fb/0x3fd
[  367.608019]  ? arch_local_irq_restore+0x53/0x53
[  367.608019]  ? vprintk_emit+0x566/0x770
[  367.608019]  ? console_unlock+0xf50/0xf50
[  367.608019]  ? kasan_check_write+0x14/0x20
[  367.608019]  ? sched_clock_local+0xe2/0x150
[  367.608019]  ? do_raw_spin_trylock+0x1a0/0x1a0
[  367.608019]  ? sched_clock_cpu+0x12e/0x170
[  367.608019]  ? memcpy+0x45/0x50
[  367.608019]  ? vprintk_emit+0x566/0x770
[  367.608019]  ? console_unlock+0xf50/0xf50
[  367.608019]  ? percpu_ref_kill_and_confirm+0xeb/0x500
[  367.608019]  ? check_noncircular+0x20/0x20
[  367.608019]  ? vprintk_default+0x28/0x30
[  367.608019]  ? vprintk_func+0x47/0x90
[  367.608019]  ? printk+0xc8/0xf9
[  367.608019]  ? load_image_and_restore+0x134/0x134
[  367.608019]  ? pointer+0xac0/0xac0
[  367.608019]  panic+0x20f/0x426
[  367.608019]  ? copy_mm+0x1219/0x1219
[  367.608019]  ? percpu_ref_kill_and_confirm+0x3ff/0x500
[  367.608019]  ? vprintk_default+0x28/0x30
[  367.608019]  ? percpu_ref_kill_and_confirm+0x3ff/0x500
[  367.608019]  __warn+0x1c4/0x1e0
[  367.608019]  warn_slowpath_fmt+0xc5/0x100
[  367.608019]  ? __warn+0x1e0/0x1e0
[  367.608019]  ? depot_save_stack+0x12c/0x480
[  367.608019]  ? css_free_rcu_fn+0x1d0/0x1d0
[  367.608019]  percpu_ref_kill_and_confirm+0x3ff/0x500
[  367.608019]  ? __percpu_ref_switch_mode+0x850/0x850
[  367.608019]  ? deactivate_super+0x173/0x1b0
[  367.608019]  ? cleanup_mnt+0xb2/0x160
[  367.608019]  ? __cleanup_mnt+0x16/0x20
[  367.608019]  ? task_work_run+0x197/0x260
[  367.608019]  ? exit_to_usermode_loop+0x23b/0x2a0
[  367.608019]  ? mark_held_locks+0xaf/0x100
[  367.608019]  ? kfree+0xed/0x250
[  367.608019]  ? trace_hardirqs_on_caller+0x545/0x6f0
[  367.608019]  ? mark_held_locks+0x100/0x100
[  367.608019]  ? lock_set_class+0xc00/0xc00
[  367.608019]  ? check_same_owner+0x320/0x320
[  367.608019]  cgroup_kill_sb+0x196/0x550
[  367.608019]  ? cgroup_migrate_add_task+0xc60/0xc60
[  367.608019]  ? unregister_shrinker+0x1c1/0x2c0
[  367.608019]  ? perf_trace_mm_vmscan_writepage+0x7a0/0x7a0
[  367.608019]  ? down_write+0x8c/0x120
[  367.608019]  ? down_read+0x150/0x150
[  367.608019]  deactivate_locked_super+0x99/0xe0
[  367.608019]  deactivate_super+0x173/0x1b0
[  367.608019]  ? mount_ns+0x190/0x190
[  367.608019]  ? dput.part.25+0x2a/0x7c0
[  367.608019]  ? dput.part.25+0x176/0x7c0
[  367.608019]  ? dput.part.25+0x2a/0x7c0
[  367.608019]  cleanup_mnt+0xb2/0x160
[  367.608019]  __cleanup_mnt+0x16/0x20
[  367.608019]  task_work_run+0x197/0x260
[  367.608019]  ? task_work_cancel+0x2f0/0x2f0
[  367.608019]  ? __unwind_start+0x380/0x380
[  367.608019]  ? entry_SYSCALL_64_fastpath+0x1f/0xc2
[  367.608019]  exit_to_usermode_loop+0x23b/0x2a0
[  367.608019]  ? trace_event_raw_event_sys_exit+0x270/0x270
[  367.608019]  ? __save_stack_trace+0x7e/0xd0
[  367.608019]  syscall_return_slowpath+0x4d3/0x570
[  367.608019]  ? prepare_exit_to_usermode+0x2e0/0x2e0
[  367.608019]  ? save_stack_trace+0x16/0x20
[  367.608019]  ? save_stack+0x43/0xd0
[  367.608019]  ? kasan_slab_free+0x6f/0xb0
[  367.608019]  ? kfree+0xd3/0x250
[  367.608019]  ? SyS_mount+0xcf/0x120
[  367.608019]  ? entry_SYSCALL_64_fastpath+0x1f/0xc2
[  367.608019]  ? mntput+0x66/0x90
[  367.608019]  ? check_noncircular+0x20/0x20
[  367.608019]  ? kfree+0xed/0x250
[  367.608019]  ? entry_SYSCALL_64_fastpath+0x93/0xc2
[  367.608019]  ? trace_hardirqs_on_caller+0x545/0x6f0
[  367.608019]  ? mark_held_locks+0x100/0x100
[  367.608019]  ? check_stack_object+0x140/0x140
[  367.608019]  ? check_stack_object+0x140/0x140
[  367.608019]  ? rcu_read_lock_sched_held+0x108/0x120
[  367.608019]  ? __kmalloc_track_caller+0x40a/0x6f0
[  367.608019]  ? SyS_mount+0xcf/0x120
[  367.608019]  ? trace_hardirqs_off+0xd/0x10
[  367.608019]  ? quarantine_put+0xea/0x190
[  367.608019]  ? SyS_mount+0xcf/0x120
[  367.608019]  ? trace_hardirqs_on_thunk+0x1a/0x1c
[  367.608019]  entry_SYSCALL_64_fastpath+0xc0/0xc2
[  367.608019] RIP: 0033:0x440b39
[  367.608019] RSP: 002b:00007f86c630adb8 EFLAGS: 00000202 ORIG_RAX:
00000000000000a5
[  367.608019] RAX: ffffffffffffffec RBX: 0000000000000000 RCX: 0000000000440b39
[  367.608019] RDX: 00000000004a0f3b RSI: 00000000004a0f34 RDI: 00000000004a0f34
[  367.608019] RBP: 00007f86c630add0 R08: 0000000000000000 R09: 0000000000000000
[  367.608019] R10: 0000000000000005 R11: 0000000000000202 R12: 0000000000000000
[  367.608019] R13: 0000000000000000 R14: 00007f86c630b9c0 R15: 00007f86c630b700
[  367.608019] Kernel Offset: disabled
[  367.608019] Rebooting in 86400 seconds..
Tejun Heo March 7, 2017, 7:45 p.m. UTC | #2 
On Tue, Mar 07, 2017 at 10:11:59AM +0100, Dmitry Vyukov wrote:
> No, still happens. Please run the repro.

Oh, I did run it but it didn't trigger the failure here.  Can you
please let me know the kernel version, config, system setup and usual
trigger duration?

Thanks.
diff mbox

Patch

diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 0125589..9c40421 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -1820,6 +1820,8 @@  static void cgroup_kill_sb(struct super_block *sb)
 	struct kernfs_root *kf_root = kernfs_root_from_sb(sb);
 	struct cgroup_root *root = cgroup_root_from_kf(kf_root);
 
+	mutex_lock(&cgroup_mutex);
+
 	/*
 	 * If @root doesn't have any mounts or children, start killing it.
 	 * This prevents new mounts by disabling percpu_ref_tryget_live().
@@ -1834,6 +1836,8 @@  static void cgroup_kill_sb(struct super_block *sb)
 		percpu_ref_kill(&root->cgrp.self.refcnt);
 
 	kernfs_kill_sb(sb);
+
+	mutex_unlock(&cgroup_mutex);
 }
 
 struct file_system_type cgroup_fs_type = {